Many AI companies fail safety measures, including Meta and OpenAI
- A report evaluated various AI companies, including OpenAI and Meta, on their safety measures.
- Meta received the lowest grade of F for its inadequate safety protocols.
- The findings indicate that AI safety measures across the industry are largely ineffective and require urgent improvement.
In a recent report by the Future of Life Institute, various AI companies were assessed regarding their safety measures in AI development. This review involved evaluations from a panel of seven independent experts, including Turing Award winner Yoshua Bengio. Companies were judged across six critical areas: risk assessment, current harms, safety frameworks, existential safety strategy, governance and accountability, and transparency and communication. The report revealed troubling vulnerabilities in flagship models from developers like OpenAI and Google DeepMind, indicating that despite claims of effective safety protocols, actual safety measures remain largely inadequate. Notably, Meta, the developer of the Llama series of AI models, received the lowest score, an F-grade, reflecting its serious shortcomings in addressing AI safety. Meanwhile, OpenAI and Google DeepMind received D+ grades, highlighting similar weaknesses in their safety frameworks. Zhipu AI received a C, indicating it has potential for improvement. The report stresses the urgency of addressing these vulnerabilities as the AI industry continues to advance, potentially outpacing the safeguards needed to ensure human control over powerful AI systems. The findings suggest that while there is significant activity within AI companies on safety-related initiatives, it is often ineffective and lacks necessary accountability. The panelists pointed out that multiple companies, including Meta and x.AI, could easily implement basic safety guidelines, yet many are not doing so. As AI systems grow in complexity, overcoming fundamental risks will necessitate advanced technical breakthroughs. The study calls for stricter adherence to safety protocols and active accountability in AI development to mitigate potential future risks posed by increasingly powerful AI technologies.